Short tip: Full vCenter Server Appliance 5.5 database prohibits logins

Recently one of my vCenter Server Appliance (vCSA) 5.5 installations was unable to proceed logins anymore. After having a look at the storage consumption I saw that two partitions were filled up:

# df -h /storage/{core,db}
Filesystem      Size  Used Avail Use% Mounted on
/dev/sdb1        20G   20G     0 100% /storage/core
/dev/sdb3        60G   56G   90M 100% /storage/db

In the VMware knowledge base there is an article regarding this symptom. This is a known issue in vCenter Server 5.5U2. The vPostgres database creates plenty of log entries like that:

WARNING: there is already a transaction in progress

The workaround is to decrease the log level. To do so, it is required to stop vCenter and the database and alter the vPostgres configuration file. Afterwards it is recommended to delete all log files except the most recent one:

# service vmware-vpxd stop
# service vmware-vpostgres stop
# cd /storage/db/vpostgres
# cp postgresql.conf postgresql.conf.orig
# vi postgresql.conf
...
312gg
...
#log_min_messages = warning
log_min_messages = error

ESC ZZ

# find . -type f -ctime +1 -exec rm {} 

One directory was cleaned – in my case there were also plenty old coredumps in the /storage/core directory. They can be removed without hesitation:

# cd /storage/core
# ls -lhs
...
2.1G -rw------- 1 root      root      2.6G Aug 18 08:46 core.vpxd-worker.25327
9.6G -rw------- 1 root      root       12G Aug 18 06:39 core.vpxd-worker.4489
# rm core.vpxd-worker.*

After starting the database and vCenter, logins should be possible again:

# service vmware-vpostgres start
# service vmware-vpxd start